Intelligent IQ classification model and perceptual ability using EEG power ratio features / Noor Hidayah Ros Azamin
Electroencephalogram (EEG) is a popular and effective approach for measuring brainwaves, as well as to explore cognitive performance like intelligence and perceptual ability. Assessment on cognitive abilities have been established via conventional psychometric tests however, the method presents a...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2020
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/61063/1/61063.pdf https://ir.uitm.edu.my/id/eprint/61063/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Electroencephalogram (EEG) is a popular and effective approach for measuring
brainwaves, as well as to explore cognitive performance like intelligence and perceptual
ability. Assessment on cognitive abilities have been established via conventional
psychometric tests however, the method presents a biasness in terms of cultural and
linguistic barriers. Previous study had implemented an intelligence quotient (IQ)
classification model via power ratio features and artificial neural network but the
findings contribute to inaccuracy of quantified features and did not achieve highest
performance measure. The cross-relational study between perception and intelligence
using EEG is relatively new subsequently, provides a new research opportunity to relate
the cognitive performance via EEG features and intelligent classification approach. This
research proposes an intelligent IQ classification model via resting EEG. It focuses on
the recorded brainwave from left prefrontal cortex where the subjects were necessarily
in relaxed state and closed eye condition. Initially, data are collected from fifty healthy
subjects and segregated into three IQ levels; low, medium and high based on Raven’s
Progressive Matrices. The brainwave features are extracted into respective bands; theta,
alpha and beta using equiripple filter and revised power ratio features. The patterns of
brainwave features are analysed for each IQ levels. The brainwave features then are
used to develop an intelligent IQ classification model. This model is implemented using
power ratio features and support vector machine with Radial Basis Function technique.
Meanwhile, the perceptual ability dataset is constructed from 65 samples which
required to complete EEG recording procedure and also perceptual ability assessment.
The samples are then segregated into perceptual ability levels; low, medium and high
based on Comprehensive Trail Making Test. The brainwave recording is further
continued with signal processing and feature extraction which is similar procedure to
IQ dataset. The extracted brainwave features are used to predict the IQ level of
perceptual ability dataset via the intelligent IQ classification model. Findings show that
intelligence and perceptual ability have positive relationship where high IQ level are
presented by high perceptual ability level and vice versa. This positive correlation was
expected by relating both attributes with attention. This study presents the intelligent IQ
classification model achieves the highest performance measure, 100% for accuracy,
sensitivity and specificity. Conclusively, this thesis proves that the brainwave features
from resting EEG present as suitable descriptors to classify and predict individual IQ
levels. The brainwave features also present as stable descriptors for perceptual ability.
Furthermore, this study confirmed that intelligence and perceptual ability can be
correlated thus, present positive relationship. |
|---|